I read a recent interview with John Batelle in which he equated the problem of search (in its more general form) with artificial intelligence. It struck me that he is quite right, and furthermore, that we will probably construct an intelligent entity as a by-product of our own desire to communicate rather than by explicitly attempting to model a more isolated and familiar form of intelligence (for example, the self-contained human brain). As such, the Turing test becomes an excellent way to think about intelligence, just not in the way we might have been led to believe. The machine which will pass the Turing test will not be isolated, it will not be the size of a human brain, and it will pass the test long before it occurs to anyone that perhaps it should be administered. We can look at the problem of search from a slightly different angle by looking at our brains. It is not about finding information in a massive database. Such a means of organizing information is very reasonable, but it does not appeal to us because that is not how we organize information in our brains (in fact, our definition of intelligence, based largely on the nature of our own brains, necessitates that any entity we would ever call intelligence store information using a method not unlike the one we employ). Thus, it is about categorization, it is about meaning, the semantic content of the information which may be be available, somewhere, to us. Finding an item in a database is easy if it is properly categorized, but we are starting to realize, with our sprawling and vast collections of information, that having a single thread, a single path which leads to a piece of information is very inconvenient. If a piece of information can be categorized in a multitude of ways, two things will occur. The first is that there will then be more possible paths which lead to the item we wish to find. The second, more subtle result is that the item in question begins to acquire meaning as a result of its categorization. This is very similar to the human brain; any single observation (for example, the combination of stimuli which lead to our notion of the defintion of a toothbrush) is meaningless without the many categorizations which can help us remember what a toothbrush is (all the times we used it, all the commercials for toothbrushes we've seen, the texture of the toothbrush, its division into parts, its shape, the string of symbols ``toothbrush," and so on). There are a myriad of ways in which we can reference something that simple, but at the same time, we must realize that these references themselves define the idea of ``toothbrush." While the second consequence is an interesting by-product, the first is most immediate in terms of motivation for improving search -- more links means more chances of getting there. To this end, the idea of tagging and cross-referencing content on an entity such as the internet grows in popularity, as do techniques like bayesian filtering, with the help of which one can avoid doing any sort of manual categorization whatsoever, allowing the entity to categorize new information using its existing experience categorizing previous items. Eventually, the goal would be to expand this to a large scale, eliminating the need to tag or categorize anything at all -- every object added to the entity should automatically be categorized using the web of meaning already established by the accumulated past observations. At that point, one can begin to see a parallel between the entity as a whole and a construct we might intuitively deem ``intelligent." We act as the sensory organs, the physical interpreters which add information to the collection of observations within the entity, while the entity itself decides how it wishes to arrange the information, how it wishes to interpret the semantic content of any new observations it receives. There is one important difference which any skeptic might enthusiastically point out: creativity, innovation. How does that fit into this? However, there is a problem we must overcome if we wish to make a claim in one direction or another -- we must define innovation. What exactly leads to innovation? Is it deterministic? Are there sufficient conditions for innovation? Are there necessary conditions? Let us assume an entity is capable of innovation. Can it innovate without having any past observations upon which it can build its ideas? I claim that it cannot. At the very least, it would need a collection of assumptions upon which (and by means of which, such as the schemes and rules of logic) it could establish new structures. Any entity within our universe is naturally imbued with some very basic rules: the laws of physics. However, in order for anything to occur, innovation aside, stimuli, information, energy must flow into the entity. It becomes evident, then, that observations, new information, are a necessary condition for innovation. I claim, also, that for an entity which tries to give meaning to observations, the observations themselves are also sufficient for creativity and innovation to characterize the actions which the entity performs. There are two points to keep in mind here. The first is the innovation and creativity inherent in the act of categorization itself. In order to categorize something new, a new observation never before observes, one must imbue the new observation with semantic content in some manner. Even if this process is deterministic, the new observation lends new meaning to existing, old observations by now sharing their ``meaning." The new observation itself has also obtained a new, original meaning. However, I've claimed that this is sufficient for true innovation. How would new ideas, new abstract constructs form as a result of merely categorizing new observations? The important, second point to keep in mind is that the cross-referenced structures which form as a result of this process are, in fact, the abstract ideas in question. Taking any subset of this collection of items and references will yield some sort of idea. However, still the entity seems somewhat static -- new observations are categorized, and any abstract idea which forms and is arguably original can be retrieved, can be ``expressed" by the entity, but can it be said that the entity ``thinks?" Unfortunately, at this point, we cannot escape our prejudices -- the best definition of consciousness and intelligence we can give is by pointing a finger confidently, without uttering a word, at a functioning human brain. It would certainly be easy to make the entity more dynamic by forcing it to constantly re- evaluate the links between its observations, making them more efficient or more robust, depending on the goal of the entity. Does the human brain do much more than this? What other motivation have we to re- evaluate our ideas, to construct new ideas without new information? It costs us quite a bit of energy to do this, and yet we find it worthwhile if it achieves some goal. Does the actual goal matter? Is it merely the constant influx of energy which necessitates this process even when a goal is not immediately present (i.e. our desire to express ourselves through art)? You may claim, then, that humans can evaluate their goals as well -- there is no reason the entity in question cannot also do the same, so long as it can observe instances or descriptions of said goals, or even parts of said goals. It is at this point that we return to the Turing test. Surely there will come a time, if such an entity is successfully constructed, when we can ask it to retrieve the information we would like (in fact, we ask for its interpretation of the information; this is already the case, a query on any popular search engine in effect asks for that engine's interpretation of the supplied query) using natural language. Indeed, we can easily do so already, by typing our request in the form of a query to some search engine. And when it replies with its interpretation of our query, it is offering us an idea, it is, in fact, engaging with us in an exchange of ideas. Surely, with enough experience and a sufficiently robust collection of cross-references, the ideas it presents can at some point rival and potentially surpass the ideas put forth by any individual human. Because judging this is difficult so long as our understand of the human brain is so poor (and I suspect our understand of the human brain will only improve after such an entity exists), the Turing test in some way becomes our only means of measuring the quality of this entity, the amount of ``intelligence" it may possess.